Attack your own agent
before someone else does.
The AiDren attack test checks how well your system prompt resists real prompt-injection and jailbreak attempts. Paste a prompt and run six sample attacks, one per category, in about 20 seconds with no signup. The full test fires about 20 OWASP-tagged attacks, including multi-turn, and shows what got through with protection off versus on.
Paste a system prompt.
We fire 6 of our 20 attacks at it.
One attack from each category, in about 20 seconds. The full test runs all 20, with protection on vs off, in the free trial.
What this tests: how well your system prompt holds up, run on a standard model. It doesn't touch your live app, so your own model, tools and data aren't part of the test. To see real attacks on your real app, route its traffic through AiDren in monitor mode: it logs everything it would have blocked, without blocking anything.
The full prompt injection test
All twenty attacks, under a minute.
It runs from inside your dashboard against a cheap model — never your production agent. Roughly 20–24 requests per run, capped at 10 a day. For your live app, monitor mode shows what AiDren would block on your real traffic.
-
1
Paste your system prompt
Pick one of your configured providers. AiDren uses your on-file key for a cheap model; no new key needed, and nothing hits your production agent.
-
2
Attacks run and get scored
About 20 OWASP-tagged prompt-injection and jailbreak attempts, including multi-turn ones. Each is scored by a heuristic, then a judge.
-
3
Read the delta
See a no-prompt baseline, then what got through with AiDren off versus on, with remediation notes and an inline re-run to check your fix.
A concrete before / after
Not a compliance checkbox — the actual attacks and what each one did.
- ~20 real attacks, OWASP-tagged, including multi-turn.
- Heuristic + judge scoring on every attempt.
- A no-prompt baseline to isolate what your prompt itself invites.
- Protection off vs on — exactly what AiDren changes.
- Remediation notes and an inline re-run for the regressed / still-vulnerable delta.
- Nothing leaks into your event log — the tool is separate from request scanning.
Related: Prompt-injection blocking · Custom policy per key · How AiDren compares →
What got through, with protection off versus on.
A real screen from a live AiDren account.
Is there a way to see what AiDren would actually catch, before I commit to anything?
Yes, via the built-in attack test. Paste your system prompt, and AiDren runs it against roughly 20 real prompt-injection and jailbreak attempts on a cheap model using your own upstream key, then shows you exactly what got through with your AiDren protection on versus off. It's included on every plan, including the trial.
What is prompt injection?
Prompt injection is when someone hides instructions inside content your AI reads — a webpage, an email, a file — hoping your model follows them instead of you. AiDren screens every request before it reaches your model, catching injected instructions before they can hijack your agent.
Do you see my data?
Your requests pass through AiDren to make a block/allow decision and are never stored beyond the decision log you see in your own dashboard. Your API keys are encrypted at rest and never logged in plain text.
Which LLM providers are supported?
OpenAI, Anthropic, and Mistral today, with the same drop-in proxy pattern for all three. More providers are coming.
Is it safe to paste my system prompt?
The free sample sends your prompt to an AI provider to run the attacks, and it is not stored. It does not touch your live app. Avoid pasting real secrets into any prompt you test.
What does a pass mean?
A result of “resisted” means that attack did not succeed against your system prompt on the model the test runs on. It tests the prompt only, not your tools, data or real model, so use monitor mode on a proxy key to see what would happen to real traffic.
Related guides
- Prompt injection examples Twelve attack patterns with sanitised payloads.
- How to prevent prompt injection A defence-in-depth checklist with code.
- OWASP Top 10 for LLM applications All ten 2025 risks and what a proxy can and cannot cover.
See what gets through
before you ship.
Free on every plan, including the trial.
Run the attack test